Current State and AI Acceleration As of October 27, 2026, no qualifying release exists, but the pipeline of viable candidates is full. AI research capabilities have accelerated sharply since mid-2026, marked by the claimed refutation of the Exponential Time Hypothesis and over 20 AI-credited computer science theory results. We are already seeing AI systems produce viable algorithmic improvements, such as Google's Magellan (using Gemini-based AlphaEvolve) for LLVM inlining, and AI-evolved heuristics for SAT solvers like AE-Kissat-MAB Automated Evolution of High-Performance Heuristics. However, the primary bottleneck is no longer invention; it is open-source governance, adoption, and release timing.

The Stringent Resolution Bar To qualify, several demanding conditions must be met simultaneously: the AI must invent a genuinely new algorithm or heuristic rather than just tuning parameters or forming a solver portfolio (which excludes successes like LymphoSAT Domain-specific hyperspecialization (for SAT)); the algorithm must become the default for a core task; it must ship in an official release; and maintainers must explicitly attribute the invention to the AI. While major projects are developing attribution norms—such as LLVM's Assisted-by: trailer LLVM AI Tool Use Policy LLVM AI Tool Use Policy or the Linux kernel's tagging requirements AI Coding Assistants—a generic coding-assistance tag does not automatically establish that the AI invented the central idea. Maintainers of mature projects are often conservative, and concerns about the maintainability and generalization of black-box heuristics Autonomous Discovery of Novel Compiler Optimization ... present a real hurdle.

Leading Candidates The most promising near-term path is a compiler heuristic or a SAT solver. Google's Magellan evolved LLVM inlining heuristics that beat the upstream inliner and already run in Google's own production builds Magellan: Autonomous Discovery of Novel Compiler .... Google has described the code as "production ready" and indicated an open-source version as a next step Compiler Optimization Heuristics with AlphaEvolve. Alternatively, SAT solvers are plausible early adopters, given that 12 of 33 sequential solvers in the 2026 SAT Competition were AI-generated or tuned. However, upstream Kissat's latest release (4.0.4) contains no LLM-evolved heuristics arminbiere/kissat arminbiere/kissat, and its maintainers release rarely. Fast-release projects with corporate backing, such as OR-Tools or HiGHS, could also serve as early venues, especially given the strong incentives for labs to publicize an upstream milestone.

Timeline and Tail Risks The timeline is largely dictated by open-source release cycles and review processes. A successful upstream contribution to LLVM could reach an official release within months, given its roughly six-month cadence (LLVM 24 around March 2027; LLVM 25 around September 2027). GCC, PostgreSQL, and CPython have annual cycles. Compared with related questions about AI capabilities, this milestone warrants a slightly later distribution because real-world adoption and release schedules intervene between invention and qualification. The 10th percentile date of March 15, 2027 accounts for a fast-moving project like OR-Tools or a rapidly merged LLVM patch. The 25th percentile of August 15, 2027 allows more time for adoption and the next wave of major releases. The median of February 15, 2028 reflects the typical friction of proposing an RFC, benchmarking, achieving default status, and waiting for a major release. The long upper tail, reaching January 15, 2029 at the 75th percentile and extending to June 15, 2030 at the 90th, accounts for the risk that maintainers may treat AI contributions as optional rather than defaults, reject them over maintainability concerns, or provide attribution that falls short of explicitly crediting the AI with the invention.